What makes a public contract offer potentially suspect of concealing a process of corruption? This is the question that three Spanish researchers are trying to answer by creating an artificial intelligence tool capable of identifying signals that can trigger alarms. After analyzing more than 70,000 contracts, the study they signed Felix Lopez Iturriaga (University of Valladolid), David Blanco Alcántara (University of Burgos) and Ivan Pasteur Sanz (International University of La Rioja) has just been published in the scientific journal European Journal of International Management.
“We cannot detect corruption in itself, but we can detect a series of criteria or ‘red flags’ that indicate less competitiveness,” explains Iván Pastor. The determining factor, according to the researcher, is often marked by the number of applications received, as this may indicate that “there is no competition” because the criteria are “very subjective or ambiguous” or “so precise” that they de facto prevent it, by making potential interested parties doubt whether they can satisfy them or by discouraging them with the idea of a “tailor-made position » .
However, synonyms and many different ways of expressing the same thing make it difficult to study these criteria and identify the signs. After all, corruption can be very superficial for data analysis. The challenge in this area therefore lies mainly in the difficulty of consulting a large volume, which is why Pastor, López and Blanco Alcántara have developed a particular methodology for their alert tool. It consists of forming four major groups conceptual, “one in which price is the almost sole criterion” and three others defined by a mixture of conditions “such as professional activity, previous experience or the technical profile of employees”, add the researchers in a press release.
Once defined, they asked the IA to review the award requirements (the contract requirements in the pre-tendering phase) and They “trained” it with a sample of over 70,000 contracts from 33 countries different published in Europe between 2016 and 2018.
One of the main conclusions reached by the researchers is that contracts awarded on the basis of price alone present a greater risk of restricting competition (most of them receive fewer than 3 offers), which increases the probability of corruption. On the other hand, contracts that evaluate aspects such as technical quality or environmental criteria tend to be more transparent and receive a greater number of applications.
Projection
“About 25 percent of the contracts analyzed received fewer than three candidates,” says Pastor, who adds that while this is not necessarily indicative of corruption, it could be. And if it’s not malicious, it usually shows, at the very least, that the offer isn’t well worded. “At present, if we introduce the characteristics of a contract into the program, it would be able to say whether or not it will have many candidates, based on historical data,” simplifies the researcher, who considers that this tool could be interesting for public administrations only to ensure its transparency, but also to ensure the effectiveness of your competitions.
However, for now, the program will not be available to institutions or businesses, says Pastor. “We want to continue research, for now we stick to the academic field and non-profit,” he says over the phone. However, he welcomes the fact that there has been progress since the “document” published in January and that this new step has served to “unite European efforts” and to arouse the interest of Transparency International, the Anti- corruption or other researchers on the subject. continent.
